Cooling LEC - Energy-flexible buildings by controlling cooling systems via unidirectional communication in local energy communities
Short Description
Progressive climate change is leading to higher summer temperatures and therefore an increased need for cooling. While the additional electricity demand from these cooling appliances is a challenge, the appliances also offer flexibility potential. Targeted utilisation could reduce the load on the electricity grids by absorbing surplus renewable generation. Control by the grid operator would make sense from a system perspective but poses the challenge of data exchange between the grid operator and the cooling appliances. Practical experience shows that this challenge is currently almost impossible to overcome.
The following research questions were therefore addressed as part of Cooling LEC:
- Is it possible to implement a control system based on unidirectional communication via ripple control systems with the aid of a self-learning system for analysing and controlling the cooling loads?
- Does the new model of energy communities offer a sufficiently economically attractive environment to be able to operate the system economically?
- How do users react to such a system?
Based on these questions, a self-learning system that uses unidirectional signalling to centrally control decentralised cooling units was developed. The aim of the control system was to optimise the time of use of the cooling units in order to use as much excess PV power as possible without compromising the comfort of the users. The Cooling LEC system consists of the Cooling LEC System Analyser, for identifying the switch-on processes of cooling units, the Cooling LEC Control Optimiser, for forecasting loads, generation, and the time of use of the cooling units and the subsequent optimisation of the switch-on times of the cooling units, and the Cooling LEC Communication Suite, for controlling the cooling units by means of a ripple control signal.
The edge detection method was used for the Cooling LEC System Analyser. Thanks to a revision of the measurement data from the cooling units to train the self-learning system, 80 % of the switch-on edges were recognised correctly during the system test. In demonstration mode, the performance of the edge detection decreased significantly and only around 40 % of the edges were recognised correctly.
For the Cooling LEC Control Optimiser, a multi-density layer model was used that predicts the use of the cooling units, the building load and the PV generation. The approach proved to be a good solution for this task, although the database for training the models was poor due to the coronavirus. To optimise the time-of-use, the approach developed in the Hybrid-Flex research project was used to optimise water heating, which optimises the use of flexibilities within a comfort window. During the system test, a root mean square error (RMSE) of around 0.2 was reached as the mean value for the cooling load forecast in both buildings. An RMSE of 0.05 was achieved for the total load forecast. In demonstration operation, the total load forecast was evaluated with the Mean Average Percentile Error (MAPE), whereby values of 60 % were achieved for the "Planning Office" demonstration building and values of approx. 23 % for the "Future House" demonstration building. The latter represents a good result.
For the evaluation of the cooling load forecast, the accuracy of the forecast was calculated, which reached a value of around 30% in the first month of the demonstration phase, but this was also due to the fact that the forecast values were created before the switch-on times were optimised. In the two following months, the forecast became significantly worse, as no training values were available for the months of September and October.
The users responded well to the switch-on recommendations in one of the two demonstration buildings, but their responsiveness declined over the course of the demonstration phase.
For the transmission of the control signals, the plan was to adapt the existing ripple control system so that the signals from the Cooling LEC Control Optimiser could be entered and sent directly. This approach was rejected for safety reasons and the ripple control system was switched to manual programming. The cooling units were not switched on directly, as they were manually controlled units; instead, the ripple control signal activated a signalling unit at the users, who then took over control. The measures taken by the users meant that a total of 40 kWh more PV electricity could be used directly.
For the economic evaluation of the approach, the integration of the demonstration buildings into a renewable energy community was analysed and alternative business models were defined. The approach turned out to be good in principle but could not be implemented due to the legal framework conditions. The use of the Cooling LEC approach as part of an energy community proved to be a sensible approach, at least that is what the simulations showed. The economic advantage was also very low at €15 p.a. due to the very low performance of the cooling units, but it showed the fundamental potential of the approach.
The involvement of the users created trust and removed barriers to implementation, but it became clear that simply making recommendations to the users was not a sufficiently reliable method for controlling the cooling units. The proof of concept confirmed the assumption that unidirectional communication has potential for energy-flexible buildings, but larger cooling units should be considered first and foremost for control.
From the methods used, it could be concluded that there is a need for good data quality and that the use of multi-dense layer models is primarily suitable for the prediction of regular signals. This is all the more important to consider as error propagation in interlinked systems is a significant issue.
Although the potential of ripple control devices is theoretically given, the realisation is difficult. Energy communities theoretically have a potential for low-cost flexibilisation measures.
Project Partners
Project management
4ward Energy Research GmbH
Project or cooperation partners
Stadtwerke Hartberg Verwaltungs Gesellschaft m.b.H.
Contact Address
Dr. Thomas Nacht
Reininghausstraße 13A
A-8020 Graz
Tel.: +43 (664) 885 00 336
E-mail: thomas.nacht@4wardenergy.at
Web: www.4wardenergy.at